On 03/16/2016 09:18 PM, Marek Vasut wrote:

> On 03/16/2016 10:58 AM, Purna Chandra Mandal wrote:
>> On 03/15/2016 11:49 PM, Marek Vasut wrote:
>>
>>> On 03/15/2016 01:44 PM, Purna Chandra Mandal wrote:
>>>> This driver adds support of PIC32 MUSB OTG controller as dual role device.
>>>> It implements platform specific glue to reuse musb core.
>>>>
>>>> Signed-off-by: Cristian Birsan <cristian.bir...@microchip.com>
>>>> Signed-off-by: Purna Chandra Mandal <purna.man...@microchip.com>
>>> [...]
>>>
>>>> diff --git a/drivers/usb/musb-new/pic32.c b/drivers/usb/musb-new/pic32.c
>>>> new file mode 100644
>>>> index 0000000..980a971
>>>> --- /dev/null
>>>> +++ b/drivers/usb/musb-new/pic32.c
>>>> @@ -0,0 +1,294 @@
>>>> +/*
>>>> + * Microchip PIC32 MUSB "glue layer"
>>>> + *
>>>> + * Copyright (C) 2015, Microchip Technology Inc.
>>>> + *  Cristian Birsan <cristian.bir...@microchip.com>
>>>> + *  Purna Chandra Mandal <purna.man...@microchip.com>
>>>> + *
>>>> + * SPDX-License-Identifier:     GPL-2.0+
>>>> + *
>>>> + * Based on the dsps "glue layer" code.
>>>> + */
>>>> +
>>>> +#include <common.h>
>>>> +#include <linux/usb/musb.h>
>>>> +#include "linux-compat.h"
>>>> +#include "musb_core.h"
>>>> +#include "musb_uboot.h"
>>>> +
>>>> +DECLARE_GLOBAL_DATA_PTR;
>>>> +
>>>> +#define PIC32_TX_EP_MASK  0x0f            /* EP0 + 7 Tx EPs */
>>>> +#define PIC32_RX_EP_MASK  0x0e            /* 7 Rx EPs */
>>>> +
>>>> +#define MUSB_SOFTRST              0x7f
>>>> +#define  MUSB_SOFTRST_NRST        BIT(0)
>>>> +#define  MUSB_SOFTRST_NRSTX       BIT(1)
>>>> +
>>>> +#define USBCRCON          0
>>>> +#define  USBCRCON_USBWKUPEN       BIT(0)  /* Enable Wakeup Interrupt */
>>>> +#define  USBCRCON_USBRIE  BIT(1)  /* Enable Remote resume Interrupt */
>>>> +#define  USBCRCON_USBIE           BIT(2)  /* Enable USB General interrupt 
>>>> */
>>>> +#define  USBCRCON_SENDMONEN       BIT(3)  /* Enable Session End VBUS 
>>>> monitoring */
>>>> +#define  USBCRCON_BSVALMONEN      BIT(4)  /* Enable B-Device VBUS 
>>>> monitoring */
>>>> +#define  USBCRCON_ASVALMONEN      BIT(5)  /* Enable A-Device VBUS 
>>>> monitoring */
>>>> +#define  USBCRCON_VBUSMONEN       BIT(6)  /* Enable VBUS monitoring */
>>>> +#define  USBCRCON_PHYIDEN BIT(7)  /* PHY ID monitoring enable */
>>>> +#define  USBCRCON_USBIDVAL        BIT(8)  /* USB ID value */
>>>> +#define  USBCRCON_USBIDOVEN       BIT(9)  /* USB ID override enable */
>>>> +#define  USBCRCON_USBWK           BIT(24) /* USB Wakeup Status */
>>>> +#define  USBCRCON_USBRF           BIT(25) /* USB Resume Status */
>>>> +#define  USBCRCON_USBIF           BIT(26) /* USB General Interrupt Status 
>>>> */
>>>> +
>>>> +static void __iomem *musb_glue;
>>> What would happen once you make a chip with two MUSB controllers ?
>> Currently PIC32 has only one MUSB controller and only one glue reg-space.
>> Don't know how the reg-map will be in future when PIC32 will have multiple
>> MUSB controllers. Assuming that glue address map will be separate for
>> each controller we can add logic to support multiple MUSB controller.
>>
>> IMO, better if we don't assume something of the future and bloat logic.
> If you switch this to driver model, you will need to weed out all the
> static global variables anyway. Better do it now.

Thanks. Will do.

>>>> +/* pic32_musb_disable - disable HDRC */
>>>> +static void pic32_musb_disable(struct musb *musb)
>>>> +{
>>> Is there no way to shut down the MUSB on the PIC32 ?
>> There is no way to disable MUSB.
> Yet another broken chip design. Can't you put the controller into reset
> and gate the clock for it ?

USB clock can't be gated! USB controller clock is derived from peripheral bus
clock(PBCLK5) which is shared with other modules and USB Phy clock derived from
UPLL can;t be gated at all.

>>>> +}
>>>> +
>>>> +/* pic32_musb_enable - enable HDRC */
>>>> +static int pic32_musb_enable(struct musb *musb)
>>>> +{
>>>> +  /* soft reset by NRSTx */
>>>> +  musb_writeb(musb->mregs, MUSB_SOFTRST, MUSB_SOFTRST_NRSTX);
>>>> +  /* set mode */
>>>> +  musb_platform_set_mode(musb, musb->board_mode);
>>>> +
>>>> +  return 0;
>>>> +}
>>>> +
>>>> +static irqreturn_t pic32_interrupt(int irq, void *hci)
>>>> +{
>>>> +  struct musb  *musb = hci;
>>>> +  irqreturn_t ret = IRQ_NONE;
>>>> +  u32 epintr, usbintr;
>>>> +
>>>> +  /* Get usb core interrupts */
>>> You mean "get" or "ack" here ?
>> I meant read-and-ack. Will update comment.
> Thanks
>
>>>> +  musb->int_usb = musb_readb(musb->mregs, MUSB_INTRUSB);
>>>> +  if (musb->int_usb)
>>>> +          musb_writeb(musb->mregs, MUSB_INTRUSB, musb->int_usb);
>>>> +
>>>> +  /* Get endpoint interrupts */
>>> DTTO
>> I meant read-and-ack.
>>
>>>> +  musb->int_rx = musb_readw(musb->mregs, MUSB_INTRRX) & PIC32_RX_EP_MASK;
>>>> +  if (musb->int_rx)
>>>> +          musb_writew(musb->mregs, MUSB_INTRRX, musb->int_rx);
>>> Same here
>> ack. Will update comment.
>>
>>>> +  musb->int_tx = musb_readw(musb->mregs, MUSB_INTRTX) & PIC32_TX_EP_MASK;
>>>> +  if (musb->int_tx)
>>>> +          musb_writew(musb->mregs, MUSB_INTRTX, musb->int_tx);
>>>> +
>>>> +  /* Drop spurious RX and TX if device is disconnected */
>>>> +  if (musb->int_usb & MUSB_INTR_DISCONNECT) {
>>>> +          musb->int_tx = 0;
>>>> +          musb->int_rx = 0;
>>>> +  }
>>>> +
>>>> +  if (musb->int_tx || musb->int_rx || musb->int_usb)
>>>> +          ret |= musb_interrupt(musb);
>>>> +
>>>> +  return ret;
>>>> +}
>>>> +
>>>> +static int pic32_musb_set_mode(struct musb *musb, u8 mode)
>>>> +{
>>>> +  struct device *dev = musb->controller;
>>>> +
>>>> +  switch (mode) {
>>>> +  case MUSB_HOST:
>>>> +          clrsetbits_le32(musb_glue + USBCRCON,
>>>> +                          USBCRCON_USBIDVAL, USBCRCON_USBIDOVEN);
>>>> +          break;
>>>> +  case MUSB_PERIPHERAL:
>>>> +          setbits_le32(musb_glue + USBCRCON,
>>>> +                       USBCRCON_USBIDVAL | USBCRCON_USBIDOVEN);
>>>> +          break;
>>>> +  case MUSB_OTG:
>>>> +          dev_err(dev, "MUSB OTG mode enabled\n");
>>> So having the core in OTG mode is an error ? Why ?
>> In PIC32 we haven't tested OTG. We use the controller as dual-role not as 
>> OTG.
>> In future we might enable that. 
> Then you might want to say "support for OTG is unimplemented" ?

Will update.

>>>> +          break;
>>>> +  default:
>>>> +          dev_err(dev, "unsupported mode %d\n", mode);
>>>> +          return -EINVAL;
>>>> +  }
>>>> +
>>>> +  return 0;
>>>> +}
>>>> +
>>>> +static int pic32_musb_init(struct musb *musb)
>>>> +{
>>>> +  u32 ctrl, hwvers;
>>>> +  u8 power;
>>>> +
>>>> +  /* Returns zero if not clocked */
>>>> +  hwvers = musb_read_hwvers(musb->mregs);
>>>> +  if (!hwvers)
>>>> +          return -ENODEV;
>>>> +
>>>> +  /* Reset the musb */
>>>> +  power = musb_readb(musb->mregs, MUSB_POWER);
>>>> +  power = power | MUSB_POWER_RESET;
>>>> +  musb_writeb(musb->mregs, MUSB_POWER, power);
>>>> +  mdelay(100);
>>>> +
>>>> +  /* Start the on-chip PHY and its PLL. */
>>>> +  power = power & ~MUSB_POWER_RESET;
>>>> +  musb_writeb(musb->mregs, MUSB_POWER, power);
>>>> +
>>>> +  musb->isr = pic32_interrupt;
>>>> +
>>>> +  ctrl =  USBCRCON_USBIF | USBCRCON_USBRF |
>>>> +          USBCRCON_USBWK | USBCRCON_USBIDOVEN |
>>>> +          USBCRCON_PHYIDEN | USBCRCON_USBIE |
>>>> +          USBCRCON_USBRIE | USBCRCON_USBWKUPEN |
>>>> +          USBCRCON_VBUSMONEN;
>>>> +  writel(ctrl, musb_glue + USBCRCON);
>>>> +
>>>> +  return 0;
>>>> +}
>>>> +
>>>> +/* PIC32 supports only 32bit read operation */
>>>> +void musb_read_fifo(struct musb_hw_ep *hw_ep, u16 len, u8 *dst)
>>>> +{
>>>> +  void __iomem *fifo = hw_ep->fifo;
>>>> +  u32 val;
>>>> +  int i;
>>> This could become:
>>>  bulk_len = len / 4;
>>>  bulk_rem = len % 4;
>>>  readsl(fifo, dst, bulk_len);
>>>  if (rem) {
>>>   dst += len & ~0x3;
>>>   tmp = readl(fifo);
>>>   copy the remaining bytes according to endianness
>>>  }
>>>
>>> This is 10 LoC , not 50 ;-)
>> May be comments in my code are not very clear :)
>>
>> All the extra code is for handling (4-byte-word) alignment of destination 
>> buffer.
>> Writing word to unaligned address will generate exception in MIPS which is 
>> not handled
>> in U-Boot software.
>> Note MIPS can't handle unaligned access in h/w unless specific unaligned
>> instructions are used.
> Ouch, did you ever see unaligned access here ?

Not seen unaligned access here but not sure of!
Char array can be starting from any arbitrary address unless it is
malloc()'d or special care is taken to make it aligned at word boundary.

I will take your logic as I'm not facing real problem.

>>>> +  /* Read for 32bit-aligned destination address */
>>>> +  if (likely((0x03 & (unsigned long)dst) == 0) && len >= 4) {
>>>> +          readsl(fifo, dst, len / 4);
>>>> +          dst += len & ~0x03;
>>>> +          len &= 0x03;
>>>> +  }
>>>> +
>>>> +  /*
>>>> +   * Now read the remaining 1 to 3 byte or complete length if
>>>> +   * unaligned address.
>>>> +   */
>>>> +  if (len > 4) {
>>>> +          for (i = 0; i < (len / 4); i++) {
>>>> +                  *(u32 *)dst = musb_readl(fifo, 0);
>>>> +                  dst += 4;
>>>> +          }
>>>> +          len &= 0x03;
>>>> +  }
>>>> +
>>>> +  if (len > 0) {
>>>> +          val = musb_readl(fifo, 0);
>>>> +          memcpy(dst, &val, len);
>>>> +  }
>>>> +}
>>> [...]
>>>

_______________________________________________
U-Boot mailing list
U-Boot@lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot

Reply via email to